Trademark Detection Using SIFT Features Matching

• Published in: 2015 International Conference on Computing Communication Control and Automation pp. 684 - 688
• Main Authors: Narhare, Ashwini D., Molke, Gajendra V.
• Format: Conference Proceeding
• Language: English
• Published: IEEE 01.02.2015
• Subjects: Accuracy; Computer vision; Detection of trademarks; Feature extraction; Histograms; Image matching; Performance analysis of detection; SIFT features matching algorithm; trademark visibility; Trademarks; Transforms
• DOI: 10.1109/ICCUBEA.2015.140

A large amount of investment every year made on sports. The major portion of which placement on billboards position in an around the field. Sponsors need to verify their brand has the level of visibility they accept to justify advertising budget. The current method of manual annotation is extremely tedious. In this work automatic annotation of trademark and logo is proposed. Detection is performed by matching set of feature descriptor for each trademark against the set of feature detected in each frame of video. The Scale Invariant Feature Transform (SIFT) algorithm proposed by Lowe [1] is an approach for extracting distinctive invariant features from image an improvement of the original SIFT algorithm providing more reliable feature matching for the purpose of trademark detection. For this image matching algorithm is designed. Further the system provides varies kind of statistics. Trademark duration visibility is calculated. Experimental results are provided along with analysis. Result shows that our proposed technique is efficient and effectively detects trademarks.